基于滑动窗优化与高斯滤波的 OFDR 应变测量 范围扩展方法

In recent years，optical frequency domain reflectometry has been widely applied in structural health monitoring，aerospace engineering，and industrial inspection owing to its high spatial resolution， high sensitivity，and distributed strain measurement capability. These characteristics enable precise and continuous detection of structural deformation along optical fibers， making optical frequency domain reflectometry a promising technique for distributed sensing applications. In optical frequency domain reflectometry-based distributed optical fiber sensing systems，the strain distribution along the fiber is typically demodulated through cross-correlation between the reference and measurement signals. However，under large-strain conditions，excessive fiber stretching leads to positional mismatch between the reference and measurement signals，resulting in spurious peaks in the cross-correlation function and consequently inaccurate strain demodulation. In addition，demodulation accuracy is influenced by several factors，including nonlinear tuning of the tunable laser source，correlation noise，detector intensity noise， and environmental vibration disturbances. As the spatial resolution increases，the number of data points involved in the cross-correlation decreases， leading to lower similarity between the reference and measurement signals and increased random noise，which further degrades strain demodulation accuracy. These issues collectively hinder the simultaneous realization of a large strain measurement range，high spatial resolution，and high demodulation accuracy in optical frequency domain reflectometry systems. To address these challenges，this paper proposes a strain measurement range extension method for optical frequency domain reflectometry based on sliding-window optimization and Gaussian filtering. The proposed method aims to extend the measurable strain range while maintaining high spatial resolution and improving demodulation accuracy. First，a sliding-window optimization strategy is employed to identify abnormal windows based on a decision criterion combining wavelength shift continuity and cross-correlation similarity. For these abnormal windows， a local adaptive search is conducted within the neighborhood of the reference window. Specifically，the reference window is incrementally shifted with a single-sample step，and the cross-correlation function is recalculated at each shifted position. The optimal matching position is determined according to the maximum cross-correlation peak criterion， thereby compensating for positional mismatch between the reference and measurement signals. This process significantly improves signal similarity and effectively suppresses spurious peaks in the cross-correlation results， thereby enhancing the robustness of strain demodulation under large-strain conditions. Furthermore， the one-dimensional cross-correlation results obtained from all sliding windows are reconstructed along the fiber length and arranged according to spatial position，forming a distribution of peak position，distance，and amplitude. By projecting this distribution onto the peak position-distance plane，a two-dimensional cross-correlation map is obtained，in which the horizontal axis represents the spatial position along the fiber，the vertical axis denotes the wavelength shift，and the color intensity indicates the peak amplitude. In this representation，spurious peaks appear as discrete noise components that can be effectively suppressed using Gaussian filtering，thereby improving the continuity and reliability of the correlation map. To validate the effectiveness of the proposed method，a distributed strain sensing experimental system based on optical frequency domain reflectometry is established. Experimental results demonstrate that， under a wavelength sweep range of 4.32 nm and a spatial resolution of 7.4 mm，the sliding-window optimization method extends the strain measurement range from 300 µε to 600 µε compared with the conventional position compensation method while reducing the strain root mean square error， mean absolute error，and standard deviation by 73.90%，80.26%，and 72.62%，respectively. Furthermore，by combining sliding-window optimization with Gaussian filtering，abnormal values in the strain demodulation results are effectively eliminated，and the strain measurement range is extended to 1 500 µε. Meanwhile， the root mean square error，mean absolute error，and standard deviation are further reduced by 98.72%， 97.12%，and 99.14%，respectively. The proposed method provides a novel demodulation scheme for extending the strain measurement range of optical frequency domain reflectometry systems while maintaining high spatial resolution，thereby broadening the application potential of optical frequency domain reflectometry and offering a practical solution for high-precision distributed optical fiber sensing.